The positive correlation found between the differential abundance measurements of carbon and nitrogen suggests that the two elements have a similar nucleosynthetic origin. Additionally given the relatively unevolved nature of these stars, it is unlikely that such variations could be caused by mixing of interior nuclear processed material to the stellar surface. For example, although the CNO-bicycle would lead to a nitrogen enhancement, there would also be a carbon depletion and hence an anti-correlation between these elements (Gies & Lambert 1992, Maeder & Meynet 1988). To obtain a carbon enrichment to match that of nitrogen, it would then be necessary to invoke subsequent nuclear processing by the conversion of helium into carbon. Such processes should not occur during a stellar main sequence or early post-main sequence evolutionary phase (Maeder & Meynet 1988, Schaller et al. 1992). Hence the observed carbon and nitrogen variations would appear to originate in inhomogeneities in the progenitor interstellar medium from which the stars formed. Such an abundance pattern also appears to support the hypothesis that carbon and nitrogen are generated by similar mechanisms, for example nucleosynthesis in intermediate mass stars. In contrast, oxygen evolution would appear to be decoupled from carbon and nitrogen which would suggest that it is generated by an unrelated mechanism, generally thought to be type II SNe in massive stars (see for example Matteucci & François 1989).
The two stars which appear to be in the cluster S289 (2 and 4) are found to have similar chemical compositions in oxygen, magnesium, aluminium and silicon. However, their carbon and nitrogen abundances differ by 0.3-0.4 dex. Although the abundance estimates for S289-2 are based on the observation of a single line for each element, the failure to detect other features in our high signal to noise spectral data, allows upper limits to be set on the equivalent widths of other lines. These confirm an underabundance of at least 0.3 dex in S289-2 compared to S289-4 for both elements. A similar result has been found for the cluster S285 (Rolleston et al. 1994) in which the star S285-1 has a nitrogen abundance approximately 0.4 dex higher than S285-6, and a carbon abundance approximately 0.2 dex higher. Again good agreement between the two cluster members is found for the abundances of other elements.
These results are surprising if the stars assigned to a given cluster are indeed members formed from the same interstellar material. To test this hypothesis, we have considered the stars assigned to S289 and tried to estimate an upper limit on their separation. Smartt et al. (1996b) have estimated masses from their positions on the diagrams of Maeder & Meynet (1988) and then, applying the bolometric corrections of Kurucz (1979), calculated the absolute visual magnitudes. Heliocentric distances of 7.0 and 8.6 kpc respectively were determined assuming a standard Galactic extinction law. The average of these two values is in good agreement with a distance of 7.9 kpc to the cluster determined by Moffat et al. (1979) by zero-age main-sequence fitting in the colour-magnitude diagram. However limits on the reliability of the measurement of surface gravity (calculated from the profiles of the hydrogen lines using the line broadening theory of Vidal et al. 1973) suggests that the logarithmic gravity may be as low as 4.0 dex in S289-4 giving a heliocentric distance of 9.6 kpc. Hence a realistic upper limit on the distance between the two stars is 2.6 kpc, which would be consistent with the observed abundance differences. However we note that for both clusters, S285 and S289, the proposed members are closely associated with an H II region, and have radial velocities and reddenings that are consistent with membership. Hence their different nitrogen and carbon abundances remain puzzling.
We have presented above differential carbon abundances which trace accurately the variations amongst the programme stars. Also we have included our estimated absolute abundances in Table 2, (on the usual scale of , with ). These are the first determinations of Population I carbon abundances in areas of low metallicity in the outer Galaxy. Indeed accurate abundances of carbon in Galactic nebular studies have not been available from optical data (see Shaver et al. 1983, Vilchez & Esteban 1996); Garnett et al. (1995, 1997) have shown the necessity of using high quality HST UV spectroscopy to tie down carbon compositions in H II regions in star-forming Galaxies. Hence there is little current information in the disk with which to compare our results. We must however be careful when comparing the absolute values in Table 2 to other studies of stellar or nebular origin. Smartt & Rolleston (1997) have highlighted the difficulties in comparing inhomogeneous data sets to trace large scale abundance variations in the Galaxy, and suggest that it is essential that any study (e.g. to determine a meaningful Galactic abundance gradient of a particular element) is based upon rigorous, self-consistent analyses. Comparing the abundances listed in Table 2 of the roughly metal-normal stars (e.g. S285-6) to those found by Gies & Lambert's (1992) study of bright nearby objects, we find our absolute values significantly lower (by 0.3 - 0.6 dex). Indeed using our line formation codes, and atomic data to estimate the abundances in the stellar data presented by Gies & Lambert, we still produce systematically lower values for carbon from the two red lines (by approximately 0.3 on average). This again reinforces the arguments in Smartt & Rolleston (1997) that studies of radial and spatial abundance variations must be done in a consistent manner.
© European Southern Observatory (ESO) 1998
Online publication: March 23, 1998